Interactive seizure & neuron simulations

In a healthy brain, neurons fire in a constantly shifting balance between excitation (mostly driven by the neurotransmitter glutamate) and inhibition (mostly driven by GABA). Voltage-gated sodium, potassium, and calcium channels decide exactly when a neuron fires and when it resets, and a layer of inhibitory interneurons acts like a crowd of circuit breakers, keeping nearby neurons from ever firing in lockstep. Zoomed out, that looks like noise on an EEG — because it is millions of independent conversations happening at once, not one shout.

A seizure starts when a group of neurons becomes abnormally excitable — in my case, inside the left temporal lobe. Scar tissue, channel dysfunction, or a local drop in inhibition can trigger what's called a paroxysmal depolarization shift: a surge of calcium and NMDA-receptor-driven depolarization that fires a neuron far harder and longer than normal, followed by a sharp after-hyperpolarization as potassium and GABA-B currents try to shut it back down. One neuron misfiring like that isn't a seizure yet. It becomes one the moment that shutdown fails and the surrounding tissue gets pulled into firing on the same beat instead of resisting it.

That's the actual mechanism: hypersynchronization. Instead of neurons firing independently, a whole region locks into one rhythm — like a crowd that starts clapping in unison instead of talking over each other. If the discharge stays contained to where it started, that's a focal seizure. If it spreads — through local cortical wiring, across the corpus callosum to the other hemisphere, or through the thalamus, which normally times rhythms across the entire cortex — it becomes a focal-to-bilateral seizure, and the clapping becomes the whole stadium.

A seizure doesn't just run out of fuel. It gets actively shut down by the same inhibitory systems that failed to stop it from starting — GABA-B receptors, potassium channels, and a flood of adenosine that acts like an emergency brake on the whole network. That's the actual mechanism behind the postictal state: the brain swings hard from hyperactive to suppressed, and the confusion, exhaustion, and memory gaps afterward are what that reset looks like from the inside.

Every major treatment I've had goes after a different point in that same chain. The 2017 lobectomy removed the actual tissue where the paroxysmal shift kept starting — cutting out the spark. The VNS, implanted in 2019, doesn't touch brain tissue directly; it stimulates the vagus nerve, which signals into the brainstem and raises inhibitory tone broadly across the cortex, making the next spark harder to catch. The DBS leads sit deep in the thalamus — the hub epilepsy devices typically target — interrupting the relay a focal discharge needs to recruit the rest of the brain into its rhythm. Three different treatments. Three different points on the exact same circuit.

The grid above shows neurons as simple on/off dots. A real neuron isn't on/off — its voltage is governed by sodium, potassium, and leak channels opening and closing continuously. In 1952, Alan Hodgkin and Andrew Huxley measured exactly how, in squid giant axon, and wrote it as four coupled differential equations. It's still the equations every computational neuroscience model builds on, and it's what's actually running below — not a canned animation, a live numerical simulation integrated in your browser via 4th-order Runge-Kutta, 60 times a second.

Turn up injected current and watch the membrane cross threshold and fire — a textbook action potential, over and over, exactly like the real thing. Then turn up inhibitory tone — the same GABA-receptor conductance that VNS and DBS lean on — and watch it shut the same neuron down, even under heavy current. This is the one-neuron version of everything described above.